STAT 100 Harder Visual Midterm Study Guide Flashcards

Causal factor

The variable that researchers hypothesize may change something else; also referred to as an exposure, treatment, predictor, or explanatory variable ($X$).

Outcome

The variable that is measured in a study to see if it changes as a result of a cause; also referred to as a response variable ($Y$).

Counterfactual

The theoretical outcome for the exact same observation under the opposite treatment condition.

Internal validity

The evaluation of whether a treatment-control comparison was fair enough to support a valid causal inference.

Baseline characteristics

Traits or conditions measured from before the treatment is administered, which can introduce causal bias if they also affect the outcome.

Case-control

A study design where units are matched on a person-to-person basis.

Cohort-control

A study design where comparisons are made whole group-to-whole group, such as comparing entire classes, schools, hospitals, or universities.

Average Treatment Effect (ATE)

A measure of causal impact that always begins with treatment minus control; calculated by averaging pair differences in matched pairs or subtracting the control group mean from the treatment group mean.

Confounder

An extraneous variable that is both related to or different across treatment groups and related to the outcome.

Adjusted effect

The estimated causal effect after removing bias, calculated as $\text{Unadjusted effect} - \text{Confounding effect}$.

Intercept

The expected or predicted value of the response variable ($Y$) when the explanatory variable ($X$) equals $0$.

Slope

The expected change in the response variable ($Y$) for each $+1$ increase in the explanatory variable ($X$).

Reference group

The baseline or control category used in linear models when dealing with a categorical explanatory variable.

Treatment indicator

A binary model term that activates an adjustment to the prediction when an observation belongs to the treatment group rather than the reference group.

Approximate margin of error

A measure of estimation uncertainty calculated as $2 \times \text{estimation error}$, centered around the point estimate.

Absolute Difference (AD)

A comparison measure between rates defined as $\text{Treatment rate} - \text{Control rate}$.

Relative Difference (RD)

A ratio comparing two group outcomes defined as $$\frac{\text{Treatment rate